ConnectOtherAIs review rounds log on a monitor in a dark office at night: each round, its reviewers and verdict, and the findings raised

ConnectOtherAIs

Independent AI review gate for coding agents
Ask about this product
Developer toolOur product

ConnectOtherAIs

Independent AI review gate for coding agents

Your coding agent cannot see its own assumptions: it reviews the diff it just wrote with the reasoning that produced it still in context, and agrees with itself. ConnectOtherAIs (coai) hands the work to models from other vendors that never saw that reasoning. The plan is reviewed before any code is written - code review is refused until the plan passes - and the diff is reviewed after, in rounds, until the findings that matter are resolved or a person is called in. It runs inside Claude Code, Cursor and Codex over MCP, before a commit exists: no bot on your pull requests, nothing installed into the repository. Every finding, and every accept or reject with its reason, is kept in a local log - the shape of your own agent's blind spots. coai is the review gate of Dew Flow, our multi-agent development orchestrator, and is free on the VS Code Marketplace as remsoftdev.connect-other-ais. As one of our developers put it: "On my latest task, verification and bug fixing took 3 hours with coai. Without it I would have spent 7 to 8."

Ask about it Learn more

Features

  • Plan gate: code review is refused until a plan round passes
  • Reviewers: Codex CLI, Antigravity (Gemini, Claude, GPT-OSS) and local models via Ollama, vLLM or any OpenAI-compatible endpoint
  • Only Blocking and Major findings count toward the gate
  • Duplicate findings from several reviewers merged into one
  • You choose what happens when rounds run out: ask a human, continue, good enough or escalate
  • Local log of every round, finding and decision, grouped by category, role and vendor
  • Native-AOT .NET 10 MCP server over stdio - no open ports
  • Help in English, German, Spanish, Ukrainian and Russian
  • Install: VS Code Marketplace - remsoftdev.connect-other-ais

The Challenge

Product: AI code review gate

Part of: Dew Flow multi-agent orchestrator

Distribution: VS Code Marketplace, MIT licence

Works with: Claude Code, Cursor, Codex

What the product had to solve:

  • An agent agrees with itself: The model that wrote the code also judges it. A reviewer from another vendor reads it without the author's reasoning - the only kind of review that catches the author's assumptions.
  • Pull-request review comes too late: By the time a PR bot comments, the code exists. coai reviews the plan first, then the diff, inside the agent's own loop.
  • Noise buries real defects: Formatting nits and the same defect raised three times waste the agent's attention. Findings are merged and only the ones that matter hold the gate.
  • Blind spots stay invisible: Without a record, nobody knows what their agent keeps missing. Every finding and decision is kept, per vendor and per category.

Results

  • Commission calculation discrepancy caught in a payroll application
  • Tenant authorization gap closed in a multi-tenant SaaS
  • Streamed API keys no longer survive secret redaction
  • Checkout concurrency hardened over four review rounds
  • 196 of 196 findings decided in one payroll project

1,087 rounds

of review, with 19,193 findings, across 14+ repositories.

5-9%

overlap between vendors - a second vendor finds different defects, not the same ones.

3 h vs 7-8 h

verification and bug fixing on a developer's recent task, with and without coai.

Case studies from our work

Interested in This Product?

Tell us how you would use it - we will answer questions on availability, setup and custom versions.

Ask about this product
An unhandled error has occurred. Reload 🗙